Public defence in Systems and Operations Research, M.Sc. (Tech) Juho Roponen
Title of the doctoral thesis: Computational Models for Adversarial Risk Analysis and Probabilistic Scenario Planning
Doctoral student: Juho Roponen
Opponent: Professor Ali Abbas,University of Southern California, USA
Custos: Professor Ahti Salo, Aalto University School of Science, Department of Mathematics and Systems Analysis
Managing risks and uncertainties using probabilities
Especially in crisis and conflict situations, decisions inevitably must be made in the face of uncertainty. In public administration and business operations, significant decisions involve not only uncertainty but also costs, work, and far-reaching consequences. Therefore, finding a good decision option, or at least avoiding the worst ones, is crucial.
The basic principles behind methods supporting decision-making under uncertainty have remained the same for a long time. If the achievement goals can be described with a clear metric such as monetary gain, and the uncertainties associated with the decision options can be represented by probability distributions, the best decision alternative can be found only using high school mathematics. However, often in reality, determining both the benefits and probability estimates is very challenging. This dissertation develops mathematical methods for handling uncertainties related to human behavior and future developments.
When modeling the decisions of multiple individuals in a conflict setting such as war, game theory is utilized. Adversarial risk analysis, for which methods are developed in this dissertation, applies solution concepts from game theory without strong assumptions about the information available to different parties or about decision-making logic. This allows for the assessment of uncertainties and the associated probabilities based on the limited information available.
Future uncertainties are addressed in the dissertation using probability-based scenario analysis. Probabilistic cross-impact analysis is employed to assess the probabilities of scenarios. When examining complex future phenomena, such as technological development, the number of significant uncertainty factors becomes high. Furthermore, uncertainty factors, such as the cost, technical performance, and adoption rate of a new technology, are not independent of one another. If a new technology is faster and cheaper than older competitors, it will quickly become widespread. Therefore, forecast models must also consider such interdependencies. To address this, the dissertation presents cross-impact analysis methods that incorporate the probabilities and pairwise dependencies of uncertainty factors based on expert knowledge. These methods can be used to assess both risks and scenario probabilities, thereby supporting decision-making.
Thesis available for public display 10 days prior to the defence at:
Doctoral theses in the School of Science: https://aaltodoc.aalto.fi/handle/123456789/52